Last modified: 2013-09-18 18:13:30 UTC
If a page is partially translated to pt and other part is translated to pt-BR, the user should be able to get a page with both translations, where the fallback language is presented instead of the original language (usually English). This is how the MediaWiki interface works, and it would avoid users making identical translations to two similar languages which are fallback one of each other (en / en-GB, pt / pt-BR, etc).
I'm leaning towards WONTFIX for this. The "always fall back to English" scheme is predictable, getting presented with a page that has Rusyn, Ukrainian, Russian and English, or Basa Banyumasan, Javanese, Indonesian and English has in my opinion too much confusion in it. It would also mean a lot of additional code complexity at little gain, in my opinion, and documents that can potentially consist of up to 3 or 4 different languages. I'm open to arguments in favour.
(In reply to comment #1) > I'm leaning towards WONTFIX for this. The "always fall back to English" > scheme > is predictable, getting presented with a page that has Rusyn, Ukrainian, > Russian and English, or Basa Banyumasan, Javanese, Indonesian and English has > in my opinion too much confusion in it. It would also mean a lot of > additional > code complexity at little gain, in my opinion, and documents that can > potentially consist of up to 3 or 4 different languages. > > I'm open to arguments in favour. PT-PT and PT-BR are almost similar. If a reader from Brazil reads a message from Portugal, he will be able to understand. There's no "too much confusion in it", because in the portuguese Wikipedia we mix the variants, and there's no confusion. There's only one Wikipedia for the variants of portuguese, it's not the same as you told about "Javanese and Indonesian", because these two languages apparently can't be mixed, as I presume after seeing that they have two different Wikipedias for them. Portuguese from Portugal and Portuguese from Brazil can be mixed. Portugueses, Galician and Spanish, can't. These are similar, but different languages, not variants of the same.
So this is about triggering a check when a XX or XX-something string is missing, searching whether strings are available in a nother XX-* variant. Modifying the subject to make this more clear. Should the strings using a different variant have some UI indication? Otherwise the reader might just think that some translator had no clue about e.g. how PT-BR is supposed to be written.
Well, currently a page such as https://www.mediawiki.org/wiki/How_to_contribute/pt?oldid=778914 which is only 33% translated to Portuguese has no visual indication that parts of it are in English. It could have at least <SomeElement lang="en">...</SomeElement> around the English text. Similarly, this attribute could be set to the code of the variant used in a given part (<translate><!--T:123-->...</translate>) of the page. And maybe this would allow us (translators?) to add some CSS to mark the text which is not in our variant. E.g. try the code mw.util.addCSS( ':not([lang|="en"])[lang]{ background: #FFC; }' ); in the browser console when viewing the page https://en.wikipedia.org/wiki/Template:Lang